runDVboostwrapper: Core function for DVboost model training

In Liuy12/DVboost: Adaptive Variant Filtering using generalized boosted regression models

Core function for DVboost model training

Description

Building training models using subset of total SVs (Only 'DEL' type)

Usage

runDVboostwrapper(var.atr.mtx, var.ID.vec, is.known.var.vec,
  output.DIR.name, input.sample.ID, fitting.verbose = FALSE,
  min.N.known.var = 50, bySVlength = FALSE)

Arguments

var.atr.mtx	a data.frame containing subset of total SVs (Only 'DEL' type)
var.ID.vec	a vector of characters indicating IDs for each SV in var.atr.mtx
is.known.var.vec	a vector contains 0 and 1s in the same order as var.atr.mtx indicating whether a SV is known(1) or novel(0)
output.DIR.name	path to output folder where all figure files and data files will be saved
input.sample.ID	character value used as prefix for output files
fitting.verbose	logical value, if TRUE, it will print out progress and performance indicators. Default to FALSE
min.N.known.var	minimum number of known SVs needed to build training model. Default to 50
bySVlength	logical value, if true, performance will also be evaluated seperately for different categories of SV length. See details section for more

Details

When bySVlength is set to true, training performance will be evaluated seperately for 5 different categories of SV length. To be specific, the 5 SV length categories are 50-100bp, 100-500bp,500-1k, 1k-10k, >10k. For each SV length category, training performance will be evaluated and exported as figures and text files.

Value

a list contains DVboost.res which is essentially a gbm.object with several additional fields:

• fitted.values: converted to probability based on fit field

• ID: supplied IDs for SVs

• is.known.variant: 0/1 indicating whether the SV is known(1) or novel(0)

• DVboost.Q.score: Q scores for SVs

Besides, this will also save a set of figures and text files to output.DIR.name. Please see metricView for details

See Also

metricView, fitDVboostmodel, DVboostQscore

Examples

data(ExampleData, package='DVboost')
sample <- 'NA12878'
outdir <- getwd()
tmp.mtx.DEL <- ExampleData[ExampleData$SVType == 'DEL',]
truth.vec <- tmp.mtx.DEL$CNVMAP == 1 | tmp.mtx.DEL$CNVR ==1
is.semi.truth.vec <- as.numeric(truth.vec)
DVb.res <- runDVboostwrapper( var.atr.mtx = tmp.mtx.DEL, var.ID.vec = rownames(tmp.mtx.DEL),
                             is.known.var.vec = is.semi.truth.vec,
                             output.DIR.name = outdir,input.sample.ID=sample, bySVlength=FALSE)

Liuy12/DVboost documentation built on May 25, 2022, 6:17 a.m.